stats_claw::algorithms::classification::naive_bayes

Struct GaussianNbModel

pub struct GaussianNbModel { /* private fields */ }

Expand description

Fitted Gaussian Naive Bayes model.

Holds the sorted class labels, the per-class log priors, and the per-class per-feature Gaussian mean and (smoothed) variance. Construct one with gaussian_nb_fit.

Implementations§

Source §

impl GaussianNbModel

Source

pub fn classes(&self) -> &[usize]

Returns the sorted class labels, in Self::predict_log_proba column order.

Source

pub fn predict(&self, x: &[Vec<f64>]) -> Result<Vec<usize>>

Predicts the class label for every sample in x (ties break low).

§Arguments

x — rows to classify; each must have the fitted feature count.

§Returns

One predicted class label per input row, in input order.

§Errors

Error::InvalidInput if any row’s length differs from the fitted feature count.

§Examples

use stats_claw::algorithms::classification::naive_bayes::gaussian_nb_fit;

let x = vec![vec![0.0], vec![0.5], vec![9.0], vec![9.5]];
let y = vec![0, 0, 1, 1];
let model = gaussian_nb_fit(&x, &y)?;
assert_eq!(model.predict(&[vec![9.1]])?, vec![1]);

Source

pub fn predict_log_proba(&self, x: &[Vec<f64>]) -> Result<Vec<Vec<f64>>>

Predicts the normalized log posteriors for every sample in x.

Each returned row has one entry per class (in Self::classes order) whose exponentials sum to 1.

§Arguments

x — rows to score; each must have the fitted feature count.

§Returns

One log-posterior row per input row.

§Errors

Error::InvalidInput if any row’s length differs from the fitted feature count.

§Examples

use stats_claw::algorithms::classification::naive_bayes::gaussian_nb_fit;

let x = vec![vec![0.0], vec![0.5], vec![9.0], vec![9.5]];
let y = vec![0, 0, 1, 1];
let model = gaussian_nb_fit(&x, &y)?;
let lp = model.predict_log_proba(&[vec![0.1]])?;
let total: f64 = lp[0].iter().map(|v| v.exp()).sum();
assert!((total - 1.0).abs() < 1e-12, "posteriors summed to {total}");

Source

pub fn classification_result( &self, x: &[Vec<f64>], y_true: &[usize], ) -> Result<ClassificationResult>

Builds a populated ClassificationResult from predictions on (x, y_true): accuracy plus macro-averaged precision / recall / F1.

§Arguments

x — evaluation design matrix.
y_true — the true class labels, one per row of x.

§Returns

A ClassificationResult scored on (x, y_true).

§Errors

Error::InvalidInput on an x/y_true length mismatch or a row whose feature count differs from the fitted model.
Error::EmptyInput if x is empty.

§Examples

use stats_claw::algorithms::classification::naive_bayes::gaussian_nb_fit;

let x = vec![vec![0.0], vec![0.5], vec![9.0], vec![9.5]];
let y = vec![0, 0, 1, 1];
let model = gaussian_nb_fit(&x, &y)?;
let result = model.classification_result(&x, &y)?;
assert!((result.accuracy - 1.0).abs() < 1e-12, "accuracy {}", result.accuracy);